Divergent genes in potential inoculant Sinorhizobium strains are related to DNA replication, recombination, and repair

Introduction

With growing areas of infertile and degraded land taken into use, the need for well-adapted nitrogen-fixing symbiotic leguminous plants is increasing. Degraded soils are often devoid of compatible rhizobia and legume plants must therefore be inoculated with appropriate strains 1. Apart from being efficient nitrogen fixers the inoculant strains should be competitive and tolerant of diverse environments. Because of the complex requirements, no successful laboratory tests have so far been developed for the selection of strains with desired properties.

Sinorhizobium meliloti is widely used for the production of inoculants for temperate Medicago species, mainly for M. sativa. However, also many tropical plants form symbioses with Sinorhizobium species 2. There is thus a demand for Sinorhizobium inoculants possibly suited for inoculation of plants in arid area. S. meliloti Rm1021, the model strain for molecular work, is an antibiotic resistant derivative of the inoculant strain SU47 3. Rm1021 is less efficient in symbiosis than its parent strain SU47 3, 4. In 2001, the genome sequence of S. meliloti Rm1021 (synonym Ensifer meliloti Rm1021) was published 5, paving the way for microarray development. With the microarrays, differential gene expression to various stress factors were determined for Rm1021 at a genome level 6-8, but the genetic basis for stress tolerance in wild-type inoculant strains was hardly studied so far.

We hypothesized that efficient inoculant strains should carry genes varying from those of Rm1021. A Rm1021 microarray was used for comparative genome hybridisations (CGH) with three salt tolerant inoculant strains. The aim was to unravel genomic features in the wild-type strains that could be related to their symbiotic efficiency and salt tolerance. We compared the physical and the functional genome composition of the reference strain Rm1021 and the wild-type strains by determining the numbers and locations of variable, that is, divergent or putatively multiplied, genes as distributed on the Rm1021 replicons (chromosome, pSymA, pSymB), and by looking for variation in clusters of orthologous groups (COGs) and in genes expressed differentially by Rm1021 after prolonged exposure to NaCl and osmotic shock.

Materials and methods

S. meliloti AK23 (RSAM1774) isolated from M. officinalis (Kazakhstan), nodulates effectively Medicago officinalis and M. sativa 9. Sinorhizobium meliloti STM 1064 from M. minima (Algeria), nodulates effectively M. minima, M. sativa, and M. ciliaris (de Lajudie, unpublished results). S. arboris HAMBI 1552 isolated from Prosopis chilensis (Sudan) is one of the most efficient symbionts of several Acacia sp. and Prosopis sp. 2. In liquid medium, AK23 and HAMBI 1552 tolerate 600 and 750 mmol l⁻¹ NaCl, respectively 10, 11, whereas Rm1021 grows poorly on LB agar with 400 mmol l⁻¹ NaCl 12. The salt tolerance of STM 1064 was estimated by growing the strain in liquid M9 medium with 0, 600, and 800 mmol l⁻¹ NaCl. Initial OD600 was adjusted to 0.1. STM 1064 grew in 600 mmol l⁻¹ NaCl, while in 800 mmol l⁻¹ NaCl there was no detectable growth.

For microarray experiments, the strains were cultivated in TY medium 13 at 30 °C in 50 ml aliquots at 180 rpm. Rm1021 was cultivated with appropriate antibiotics (streptomycin, 600 μg ml⁻¹; nalidixic acid, 8 μg ml⁻¹; kanamycin, 200 μg ml⁻¹). Cultures were harvested after 19–22 h of incubation at OD600 = 0.8. DNA was extracted with the FastDNA Kit (MP Biomedicals) according to the manufacturer's instructions and quantified by spectrophotometry (Biophotometer, Eppendorf AG, Germany). Eight microgram of DNA was labelled by incorporation of Cy5-dCTP and Cy3-dCTP fluorescent dyes (test DNA/reference DNA (Rm1021) and vice versa) (Amersham Biosciences, UK) using the RadPrime DNA labelling kit (Invitrogen) and purified using PCR purification kit (Qiagen, Netherlands). The test and reference DNA were combined with 10 µg Herring sperm DNA, 20 µg poly-(dA) and 100 µg ribonuclease free yeast tRNA, vacuum dried, resuspended in 80 µl hybridisation buffer (50% formamide/6xSSC/0.5% SDS/5xDenhart's) and hybridized to the Sm6kOligo microarray 7 at 42 °C overnight. Arrays were scanned with GenePix Autoloader 4200A (Molecular Devices, LLC). The images were processed using GenePix 6.0 software.

Systematic variation normalization was performed using the R software 14. The divergent genes were identified using permutation-T-tests where the parameter for false discovery rate was set to zero. The probability of a gene to be randomly picked as divergent was calculated. Two deletion mutants of S. meliloti Rm1021, namely S. meliloti Rm1021 ΔsitA and S. meliloti Rm1021 Δfur, were used to optimize the CGH experiment. Both the reference strain and the test strains were labelled with both dyes and the experiment was done in duplicate. However the signal intensity obtained in the alternating dyes, particularly on the test strains was not proportional. Since swapping the dye did not produce the expected result we decided to label the reference strain with Cy5 and the test strains with Cy3 in subsequent experiments.

ImaGene 5.0 software (Biodiscovery Inc., Los Angeles, CA) was used for spot detection, image segmentation, and signal quantification. The log₂ value of the intensity ratios was calculated for each spot with M_i = log₂(R_i/G_i), where R_i = I_ch1i-Bg_ch1i and G_i = I_ch2i-Bg_ch2i with I_ch1i and I_ch2i being the intensity of a spot in channel l or channel 2, and Bg_ch1i and Bg_ch2i being the background intensity of a spot in channel 1 or channel 2, respectively. The mean intensity was calculated for each spot with A_i = log₂(RiGi) 0.5. Normalization and t-statistics were carried out using the Emma 1.1 software 15. A gene was considered significantly different if p ≤ 0.05, the log₂ ratio of the intensities (M-value) was ≥1 or ≤−1, and the mean intensity (A-value) was ≥8.

For K-means clustering analysis of the microarray data, the Genesis© software was used 16. Clustering with TMEV (version 3.1) software revealed that the replicate slides behaved in almost identical manner. The MA plot generated after the T-permutation test also confirmed the clustering. We calculated whether the distribution of divergent genes on different replicons, in different COG categories, and in genes affected by prolonged cultivation with 380 mmol l⁻¹ NaCl 8 and by osmotic shock 6 were significant in comparison with the total number of genes with the Chi-square (χ²) test. The COG categories are as listed at Galardini et al. 17 (http://www.genomebiology.com/content/supplementary/1471-2164-12-235-s2.docx).

Results and discussion

To search for genetic determinants for effective nodulation and salt stress tolerance, we compared the genomes of three salt tolerant inoculant Sinorhizobium strains to the genome of S. meliloti Rm1021. In earlier studies, the number of variable genes in wild-type S. meliloti strains ranged from 170 to 653 18, 19. The variation was connected neither to original host nor to site of isolation 19. In our study, S. arboris HAMBI 1552 had more variable genes than the phylogenetically more closely related S. meliloti strains AK23 and STM 1064 (Table 1). It should be noted that CGH gives information only on the genes present in the Rm1021 genome. As in earlier studies 18, 19, the majority of the variable genes of the wild type strains showed decreased hybridisation intensity indicating gene deletion or sequence divergence (referred as divergent genes hereafter) in the probe region (Table 1).

For further insight into the genomic variation, we determined the location of the variable genes on the three genetic elements of Rm1021. In general, the housekeeping genes are on the chromosome, the genes related to symbiosis on pSymA and the genes related to survival in the free-living state in soil on pSymB 5. The genes that show increased hybridisation intensity are considered putatively duplicated 20. In S. arboris HAMBI 1552, the putatively duplicated genes were more abundant on the chromosome than on other genetic elements (Table 1). As in earlier studies 18, 19, 21, in all the strains the divergent genes were more abundant on pSymA (Table 1). Contrary to the chromosome and pSymB, on pSymA the proportion of accessory genes is higher than that of the core genes 19. The accessory genes include genes related to environmental adaptation that are likely to explain the phenotypic variation among rhizobial strains 17, 22.

For deeper understanding of the genomic differences, the variable genes were assigned to clusters of orthologous group (COG) categories 23. Earlier studies have shown that phosphate starvation, osmotic shock, and salt stress result in differential gene expression 6-8. Considering both all the Rm1021 genes and those induced by osmotic shock 6, 8, in the wild-type strains the divergent genes were more abundant in COG category L, that is, DNA replication, recombination and repair genes (Tables 1 and 2). With the exception of genes with no assigned COG category (COG X) of STM1064 and HAMBI 1552, the relative proportion of divergent genes in other COGs reflected the overall divergence. Stress induced mutagenesis is important in the adaptation of bacteria to changing conditions 24, and COG L genes are involved in the process. For S. meliloti strains AK83 and BL225C, the genes in the COG L were the most diverse, and COG L and COG X were the only groups where the number of accessory genes was higher than that of core genes 17. In Rm1021, 20 genes in the COG L that were affected by osmotic shock code for transposases 6, out of which all but one duplicated gene were divergent in HAMBI 1552 (Table 1). Eight and nine of the ten transposase genes induced by osmotic shock were divergent in AK23 and STM 1064, respectively (Table 1). In addition, the relative proportion of divergent transposase genes in the wild type strains was 4.5–4.8 times higher than that of all the divergent genes (Table 1). Genomic rearrangement by transposition could be considered to play a major role in adaptation 17. Although we cannot conclude how the divergence among the COG L and transposase genes affects the wild type strains, the observation implies that the strains have a strategy considerably different to that of Rm1021 for responding to osmotic shock and possibly to other stress conditions. In Rm1021, 444 genes were down-regulated under osmotic stress 6. Altogether 40 COG J genes, that is, translation, ribosomal structure, and biogenesis genes, were down-regulated by osmotic shock, presumably because protein biosynthesis is less intense at the lower growth rate caused by stress 6. In all the wild-type strains, the proportion of genes divergent from the down-regulated genes was approximately half of the total divergence (Tables, 2 1), implying that these genes and possibly also the down-regulation response are conserved. Duplication of genes is often associated with functional diversification 25. Altogether 44 COG J genes, including twenty osmotic shock affected genes, were duplicated in HAMBI 1552 (Table 1), reflecting the evolutionary distance between the S. meliloti and S. arboris strains.

The long term cultivation of Rm1021 in 380 mmol l⁻¹ NaCl up-regulated 52 and down-regulated 85 genes 8. Both AK23 and STM 1064 were divergent from 12 of the down-regulated genes (Table 1). Seven of these were related to synthesis of rhizobactin that is involved in the ability of Rm1021 to survive in iron deficient soils 26. Actually, all the eight rhizobactin operon genes of Rm1021 were divergent in AK23 and STM 1064 (Table S1). Contrary to the wild-type S. meliloti strains, the genes of HAMBI 1552 variable to the salt stress affected genes of Rm1021 included equal number of multiplied and divergent genes, among others, three divergent rhizobactin genes and two multiplied iron uptake genes. The rhizobactin operon is missing from the genomes of Aral sea isolates S. meliloti AK 21 and AK83 17, 27, implying that the wild type strains have alternative, possibly more stress tolerant, systems for iron uptake.

The 22 nodulation linked genes on the Rm1021 microarray include both structural nod genes and the nodD genes which are involved in the activation of Nod factor synthesis 7, 28. The NodD1 and NodD2 in Rm1021 are responsive to different compounds in plant root secretions whereas NodD3 is plant inducer-independent 28. Both AK23 and STM 1064 had two variable nodulation genes (Table 3). The multiplication of nodD3 and Nod factor secretion-linked nodJ in AK23 and STM 1064, respectively, is possibly connected to the efficient nodulation capabilities of the strains. Rm1021 and HAMBI 1552 have different host ranges; they nodulate herbaceous legumes and tropical trees, respectively. Also, the structures of their Nod factors are different 28, 29. In line with these differences, the nodD2, the nodL connected to acetylation of the Nod factor non-reducing terminal and nodE linked to Nod factor acylation by polyunsaturated fatty acids were divergent (Table 3).

Concluding remarks

Both the physical location and COG category assignment of the divergent genes implied that the divergence was important in shaping the adaptability of the strains. Even though the microarray method successfully revealed the differences between the model strain Rm1021 and the potential inoculant strains, more knowledge on the genetic determinants of effective nodulation and stress tolerance is needed before genetic screening can replace traditional nodulation tests.

Conflict of interest

The authors have declared no conflict of interest.